Identifying optimal combination regimens for therapy of Mycobacterium tuberculosis with an algorithmic approach: prospective predictions and validations.

Abstract

Mycobacterium tuberculosis resistance to standard-of-care agents is increasing. It is imperative to identify new combinations that increase the rate and depth of bacterial kill, shorten therapy and also suppress resistance. There has been little prior effort to identify combination regimens that employ new or repurposed drugs in a rational way. Our group developed a pathway to combine agents to achieve this end. This pathway starts with standard baseline evaluations (e.g., MIC), leverages information from in vitro assessments (hollow fiber infection model), then analyzes 2-agent combinations in a 96 well quantitative culture checkerboard format (Greco URSA model with simulation). Finally, development of a high dimensional mathematical model allowed evaluation of 2- and 3-drug regimens in multiple metabolic states to draw inferences regarding combination therapies. We prospectively evaluated these regimens in animal models. We showed that a prospectively chosen regimen of pretomanid, moxifloxacin plus bedaquiline performed as predicted. In the BALB/c murine model, this regimen produced sterilization in a cohort that was held for 12 weeks after therapy cessation, as it did in the C3HeB/FeJ ("Kramnik") murine model. Finally, this and other regimens were evaluated in a cynomolgus macaque model. The decrement of the 18F-deoxyglucose signal in Positron emission tomography (PET)- computed tomography (CT) evaluations was best with this regimen. Other endpoints such as necropsy score and colony counts in lung and lymph nodes also demonstrated that this regimen behaved as predicted from our pathway/algorithm. We conclude that this provides a way forward for the future to identify the most promising regimens to shorten therapy for tuberculosis and suppress emergence of resistance.